未验证 提交 4ece8716 编写于 作者: S Shengliang Guan 提交者: GitHub

Merge pull request #4519 from taosdata/feature/wal

Feature/wal
......@@ -206,7 +206,7 @@ int32_t tsNumOfLogLines = 10000000;
int32_t mDebugFlag = 131;
int32_t sdbDebugFlag = 131;
int32_t dDebugFlag = 135;
int32_t vDebugFlag = 131;
int32_t vDebugFlag = 135;
int32_t cDebugFlag = 131;
int32_t jniDebugFlag = 131;
int32_t odbcDebugFlag = 131;
......
......@@ -161,7 +161,7 @@ void cqStop(void *handle) {
return;
}
SCqContext *pContext = handle;
cInfo("vgId:%d, stop all CQs", pContext->vgId);
cDebug("vgId:%d, stop all CQs", pContext->vgId);
if (pContext->dbConn == NULL || pContext->master == 0) return;
pthread_mutex_lock(&pContext->mutex);
......
......@@ -206,9 +206,10 @@ TAOS_DEFINE_ERROR(TSDB_CODE_VND_NO_SUCH_FILE_OR_DIR, 0, 0x0507, "Missing da
TAOS_DEFINE_ERROR(TSDB_CODE_VND_OUT_OF_MEMORY, 0, 0x0508, "Out of memory")
TAOS_DEFINE_ERROR(TSDB_CODE_VND_APP_ERROR, 0, 0x0509, "Unexpected generic error in vnode")
TAOS_DEFINE_ERROR(TSDB_CODE_VND_INVALID_VRESION_FILE, 0, 0x050A, "Invalid version file")
TAOS_DEFINE_ERROR(TSDB_CODE_VND_IS_FULL, 0, 0x050B, "Vnode memory is full because commit failed")
TAOS_DEFINE_ERROR(TSDB_CODE_VND_IS_FULL, 0, 0x050B, "Database memory is full for commit failed")
TAOS_DEFINE_ERROR(TSDB_CODE_VND_IS_FLOWCTRL, 0, 0x050C, "Database memory is full for waiting commit")
TAOS_DEFINE_ERROR(TSDB_CODE_VND_NOT_SYNCED, 0, 0x0511, "Database suspended")
TAOS_DEFINE_ERROR(TSDB_CODE_VND_NO_WRITE_AUTH, 0, 0x0512, "Write operation denied")
TAOS_DEFINE_ERROR(TSDB_CODE_VND_NO_WRITE_AUTH, 0, 0x0512, "Database write operation denied")
TAOS_DEFINE_ERROR(TSDB_CODE_VND_SYNCING, 0, 0x0513, "Database is syncing")
// tsdb
......
......@@ -282,25 +282,32 @@ static int32_t mnodeRetrieveConns(SShowObj *pShow, char *data, int32_t rows, voi
// not thread safe, need optimized
int32_t mnodeSaveQueryStreamList(SConnObj *pConn, SHeartBeatMsg *pHBMsg) {
pConn->numOfQueries = htonl(pHBMsg->numOfQueries);
if (pConn->numOfQueries > 0) {
pConn->numOfQueries = 0;
pConn->numOfStreams = 0;
int32_t numOfQueries = htonl(pHBMsg->numOfQueries);
if (numOfQueries > 0) {
if (pConn->pQueries == NULL) {
pConn->pQueries = calloc(sizeof(SQueryDesc), QUERY_STREAM_SAVE_SIZE);
}
int32_t saveSize = MIN(QUERY_STREAM_SAVE_SIZE, pConn->numOfQueries) * sizeof(SQueryDesc);
pConn->numOfQueries = MIN(QUERY_STREAM_SAVE_SIZE, numOfQueries);
int32_t saveSize = pConn->numOfQueries * sizeof(SQueryDesc);
if (saveSize > 0 && pConn->pQueries != NULL) {
memcpy(pConn->pQueries, pHBMsg->pData, saveSize);
}
}
pConn->numOfStreams = htonl(pHBMsg->numOfStreams);
if (pConn->numOfStreams > 0) {
int32_t numOfStreams = htonl(pHBMsg->numOfStreams);
if (numOfStreams > 0) {
if (pConn->pStreams == NULL) {
pConn->pStreams = calloc(sizeof(SStreamDesc), QUERY_STREAM_SAVE_SIZE);
}
int32_t saveSize = MIN(QUERY_STREAM_SAVE_SIZE, pConn->numOfStreams) * sizeof(SStreamDesc);
pConn->numOfStreams = MIN(QUERY_STREAM_SAVE_SIZE, numOfStreams);
int32_t saveSize = pConn->numOfStreams * sizeof(SStreamDesc);
if (saveSize > 0 && pConn->pStreams != NULL) {
memcpy(pConn->pStreams, pHBMsg->pData + pConn->numOfQueries * sizeof(SQueryDesc), saveSize);
}
......
......@@ -1312,7 +1312,7 @@ static int32_t syncForwardToPeerImpl(SSyncNode *pNode, void *data, void *mhandle
}
// always update version
sTrace("vgId:%d, forward to peer, replica:%d role:%s qtype:%s hver:%" PRIu64, pNode->vgId, pNode->replica,
sTrace("vgId:%d, update nodeVersion, replica:%d role:%s qtype:%s hver:%" PRIu64, pNode->vgId, pNode->replica,
syncRole[nodeRole], qtypeStr[qtype], pWalHead->version);
nodeVersion = pWalHead->version;
......
......@@ -43,7 +43,7 @@ static void syncRemoveExtraFile(SSyncPeer *pPeer, int32_t sindex, int32_t eindex
snprintf(fname, sizeof(fname), "%s/%s", pNode->path, name);
(void)remove(fname);
sDebug("%s, %s is removed", pPeer->id, fname);
sInfo("%s, %s is removed for its extra", pPeer->id, fname);
index++;
if (index > eindex) break;
......
......@@ -46,6 +46,7 @@ typedef struct {
int8_t isFull;
int8_t isCommiting;
uint64_t version; // current version
uint64_t cversion; // version while commit start
uint64_t fversion; // version on saved data file
void * wqueue; // write queue
void * qqueue; // read query queue
......
......@@ -203,8 +203,8 @@ int32_t vnodeOpen(int32_t vgId) {
code = vnodeReadVersion(pVnode);
if (code != TSDB_CODE_SUCCESS) {
vError("vgId:%d, failed to read version, generate it from data file", pVnode->vgId);
// Allow vnode start even when read version fails, set version as walVersion or zero
vError("vgId:%d, failed to read file version, generate it from data file", pVnode->vgId);
// Allow vnode start even when read file version fails, set file version as wal version or zero
// vnodeCleanUp(pVnode);
// return code;
}
......@@ -442,6 +442,7 @@ static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno) {
pVnode->fversion, pVnode->version);
pVnode->isCommiting = 0;
pVnode->isFull = 1;
pVnode->cversion = pVnode->version;
return 0;
}
......@@ -457,7 +458,7 @@ static int32_t vnodeProcessTsdbStatus(void *arg, int32_t status, int32_t eno) {
if (status == TSDB_STATUS_COMMIT_OVER) {
pVnode->isCommiting = 0;
pVnode->isFull = 0;
pVnode->fversion = pVnode->version;
pVnode->fversion = pVnode->cversion;
vDebug("vgId:%d, commit over, fver:%" PRIu64 " vver:%" PRIu64, pVnode->vgId, pVnode->fversion, pVnode->version);
if (!vnodeInInitStatus(pVnode)) {
walRemoveOneOldFile(pVnode->wal);
......
......@@ -133,7 +133,7 @@ static int32_t vnodePutItemIntoReadQueue(SVnodeObj *pVnode, void **qhandle, void
int32_t code = vnodeWriteToRQueue(pVnode, qhandle, 0, TAOS_QTYPE_QUERY, &rpcMsg);
if (code == TSDB_CODE_SUCCESS) {
vDebug("QInfo:%p add to vread queue for exec query", *qhandle);
vTrace("QInfo:%p add to vread queue for exec query", *qhandle);
}
return code;
......@@ -164,7 +164,7 @@ static int32_t vnodeDumpQueryResult(SRspRet *pRet, void *pVnode, void **handle,
}
} else {
*freeHandle = true;
vDebug("QInfo:%p exec completed, free handle:%d", *handle, *freeHandle);
vTrace("QInfo:%p exec completed, free handle:%d", *handle, *freeHandle);
}
} else {
SRetrieveTableRsp *pRsp = (SRetrieveTableRsp *)rpcMallocCont(sizeof(SRetrieveTableRsp));
......@@ -266,7 +266,7 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
}
if (handle != NULL) {
vDebug("vgId:%d, QInfo:%p, dnode query msg disposed, create qhandle and returns to app", vgId, *handle);
vTrace("vgId:%d, QInfo:%p, dnode query msg disposed, create qhandle and returns to app", vgId, *handle);
code = vnodePutItemIntoReadQueue(pVnode, handle, pRead->rpcHandle);
if (code != TSDB_CODE_SUCCESS) {
pRsp->code = code;
......@@ -278,7 +278,7 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
assert(pCont != NULL);
void **qhandle = (void **)pRead->qhandle;
vDebug("vgId:%d, QInfo:%p, dnode continues to exec query", pVnode->vgId, *qhandle);
vTrace("vgId:%d, QInfo:%p, dnode continues to exec query", pVnode->vgId, *qhandle);
// In the retrieve blocking model, only 50% CPU will be used in query processing
if (tsHalfCoresForQuery) {
......@@ -294,7 +294,7 @@ static int32_t vnodeProcessQueryMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
pRead->rpcHandle = qGetResultRetrieveMsg(*qhandle);
assert(pRead->rpcHandle != NULL);
vDebug("vgId:%d, QInfo:%p, start to build retrieval rsp after query paused, %p", pVnode->vgId, *qhandle,
vTrace("vgId:%d, QInfo:%p, start to build retrieval rsp after query paused, %p", pVnode->vgId, *qhandle,
pRead->rpcHandle);
// set the real rsp error code
......@@ -327,7 +327,7 @@ static int32_t vnodeProcessFetchMsg(SVnodeObj *pVnode, SVReadMsg *pRead) {
pRetrieve->free = htons(pRetrieve->free);
pRetrieve->qhandle = htobe64(pRetrieve->qhandle);
vDebug("vgId:%d, QInfo:%p, retrieve msg is disposed, free:%d, conn:%p", pVnode->vgId, (void *)pRetrieve->qhandle,
vTrace("vgId:%d, QInfo:%p, retrieve msg is disposed, free:%d, conn:%p", pVnode->vgId, (void *)pRetrieve->qhandle,
pRetrieve->free, pRead->rpcHandle);
memset(pRet, 0, sizeof(SRspRet));
......@@ -410,6 +410,6 @@ int32_t vnodeNotifyCurrentQhandle(void *handle, void *qhandle, int32_t vgId) {
pMsg->header.vgId = htonl(vgId);
pMsg->header.contLen = htonl(sizeof(SRetrieveTableMsg));
vDebug("QInfo:%p register qhandle to connect:%p", qhandle, handle);
vTrace("QInfo:%p register qhandle to connect:%p", qhandle, handle);
return rpcReportProgress(handle, (char *)pMsg, sizeof(SRetrieveTableMsg));
}
......@@ -244,7 +244,7 @@ int32_t vnodeWriteToWQueue(void *vparam, void *wparam, int32_t qtype, void *rpar
int32_t queued = atomic_add_fetch_32(&pVnode->queuedWMsg, 1);
if (queued > MAX_QUEUED_MSG_NUM) {
vDebug("vgId:%d, too many msg:%d in vwqueue, flow control", pVnode->vgId, queued);
taosMsleep(1);
taosMsleep(3);
}
code = vnodePerformFlowCtrl(pWrite);
......@@ -271,6 +271,8 @@ static void vnodeFlowCtrlMsgToWQueue(void *param, void *tmrId) {
SVnodeObj * pVnode = pWrite->pVnode;
int32_t code = TSDB_CODE_VND_SYNCING;
if (pVnode->flowctrlLevel <= 0) code = TSDB_CODE_VND_IS_FLOWCTRL;
pWrite->processedCount++;
if (pWrite->processedCount > 100) {
vError("vgId:%d, msg:%p, failed to process since %s, retry:%d", pVnode->vgId, pWrite, tstrerror(code),
......@@ -290,8 +292,8 @@ static void vnodeFlowCtrlMsgToWQueue(void *param, void *tmrId) {
static int32_t vnodePerformFlowCtrl(SVWriteMsg *pWrite) {
SVnodeObj *pVnode = pWrite->pVnode;
if (pVnode->flowctrlLevel <= 0) return 0;
if (pWrite->qtype != TAOS_QTYPE_RPC) return 0;
if (pVnode->queuedWMsg < MAX_QUEUED_MSG_NUM && pVnode->flowctrlLevel <= 0) return 0;
if (tsFlowCtrl == 0) {
int32_t ms = pow(2, pVnode->flowctrlLevel + 2);
......
......@@ -38,7 +38,7 @@ extern int32_t wDebugFlag;
#define WAL_SIGNATURE ((uint32_t)(0xFAFBFDFE))
#define WAL_PATH_LEN (TSDB_FILENAME_LEN + 12)
#define WAL_FILE_LEN (WAL_PATH_LEN + 32)
#define WAL_FILE_NUM 3
#define WAL_FILE_NUM 1 // 3
typedef struct {
uint64_t version;
......
......@@ -173,7 +173,7 @@ int32_t walRestore(void *handle, void *pVnode, FWalWrite writeFp) {
continue;
}
wInfo("vgId:%d, file:%s, restore success", pWal->vgId, walName);
wInfo("vgId:%d, file:%s, restore success, wver:%" PRIu64, pWal->vgId, walName, pWal->version);
count++;
}
......@@ -267,8 +267,6 @@ static int32_t walRestoreWalFile(SWal *pWal, void *pVnode, FWalWrite writeFp, ch
return TAOS_SYSTEM_ERROR(errno);
}
wDebug("vgId:%d, file:%s, start to restore", pWal->vgId, name);
int32_t code = TSDB_CODE_SUCCESS;
int64_t offset = 0;
SWalHead *pHead = buffer;
......
......@@ -91,8 +91,11 @@ while $i < $tblNum
$i = $i + 1
endw
sql show db.vgroups;
print d1: $data04 $data05 , d2: $data06 $data07
sql select count(*) from $stb
print rows:$rows data00:$data00
print rtest1==> rows:$rows data00:$data00
if $rows != 1 then
return -1
endi
......@@ -103,6 +106,15 @@ endi
$totalRows = $data00
sql select count(*) from $stb
print test2==> rows:$rows data00:$data00
sql select count(*) from $stb
print test3==> rows:$rows data00:$data00
sql select count(*) from $stb
print test4==> rows:$rows data00:$data00
sql select count(*) from $stb
print test5==> rows:$rows data00:$data00
print ============== step3: insert old data(now-15d) and new data(now+15d), control data rows in order to save in cache, not falling disc
sql insert into $tb values ( now - 20d , -20 )
sql insert into $tb values ( now - 40d , -40 )
......@@ -153,12 +165,21 @@ if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
sql select count(*) from $stb
print data00 $data00
sql select count(*) from $stb
print data00 $data00
sql select count(*) from $stb
print data00 $data00
print ============== step5: insert two data rows: now-16d, now+16d,
sql insert into $tb values ( now - 21d , -21 )
sql insert into $tb values ( now - 41d , -41 )
$totalRows = $totalRows + 2
print ============== step5: restart dnode2, waiting sync end
print ============== step6: restart dnode2, waiting sync end
system sh/exec.sh -n dnode2 -s start
sleep 3000
$loopCnt = 0
......@@ -192,9 +213,81 @@ endi
sleep $sleepTimer
# check using select
sleep 5000
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
sql select count(*) from $stb
print data00 $data00
if $data00 != $totalRows then
return -1
endi
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册